On Two Challenging Problems in Statistical Debugging
Title: On Two Challenging Problems in Statistical Debugging
Speaker: Xiaoyuan Xie (Swinburne University of Technology, Australia)
Time: 14:30, August 30th, 2013
Venue: Lecture Room, 3rd Floor, Building #5, State Key Laboratory of Computer Science, Institute of Software, Chinese Academy of Sciences
Abstract:
Statistical debugging is one of the most popular debugging approaches, in which data can be systematically collected to predict which statement is faulty. Spectrum-based fault localization (SBFL), as a big family of statistical debugging approaches, has received much attention in these years. For the past decade, over hundreds of SBFL papers were published. However, there are two challenging problems. One problem is that all studies have assumed the existence of a test oracle. However, in many application domains, this assumption does not hold. We use the technique of metamorphic testing to address this problem. Another problem is that all previous studies used experimental analysis to evaluate various SBFL techniques. However, results of these analysis cannot be conclusive because their validity is limited only to those subject programs that are investigated. To address this problem, we have proposed a framework, which can theoretically evaluate a SBFL technique.
Statistical debugging is one of the most popular debugging approaches, in which data can be systematically collected to predict which statement is faulty. Spectrum-based fault localization (SBFL), as a big family of statistical debugging approaches, has received much attention in these years. For the past decade, over hundreds of SBFL papers were published. However, there are two challenging problems. One problem is that all studies have assumed the existence of a test oracle. However, in many application domains, this assumption does not hold. We use the technique of metamorphic testing to address this problem. Another problem is that all previous studies used experimental analysis to evaluate various SBFL techniques. However, results of these analysis cannot be conclusive because their validity is limited only to those subject programs that are investigated. To address this problem, we have proposed a framework, which can theoretically evaluate a SBFL technique.
Biography: